This audit evaluated patients treated in the clinic from 01 January 2024 to 21 May 2025, with an initial visit restricted to dates occurring before 21 September 2024, to ensure a minimum available follow-up of 8 months. Figure 1 displays the patient selection process, highlighting the exclusion criteria.

Two physicians were responsible for patient interaction and implementation of the dietary intervention. All patients of any age (> 18 years) were included in the study, provided they met all inclusion criteria. For all included patients, all laboratory values within the study period were collected, provided they were ordered after establishing care at the clinic of interest. Patients were not excluded based on a history of mental health or eating disorders.

At the initial visit, patients are scheduled for a 1-h consultation designed to address their chief complaints, review current dietary habits and discuss the VLCD implementation. Patients are given a food questionnaire to assess their current eating patterns (see Online Appendix 1), which guides the physicians in how they want to implement lifestyle changes for each individual patient. The first steps include examining what the patient’s nutrition currently looks like, specifically their carbohydrate intake. It is advised that patients decrease their overall carbohydrate intake. There is a nutritional difference in whole food carbohydrates, such as vegetables and fruit, and processed carbohydrates, such as breads and cereals. Individuals could be healthy at a higher carbohydrate intake than recommended in this clinic; however, because of the patient population often presenting with signs of metabolic dysregulation, carbohydrates from all sources are recommended to be reduced. Initial advice is to include more meat and egg consumption, as this increases protein intake in place of carbohydrate sources. Patients are also advised to avoid sugar-sweetened beverages if they are currently a part of their diet. These dietary discussions are tailored to each patient; however, these are readily accessible changes that are addressed with most patients.

Laboratory values are also often obtained; high-density lipoprotein cholesterol (HDL-c), triglycerides (TGs), haemoglobin A1c (HbA1c), high-sensitivity C-reactive protein (hsCRP), fasting glucose and insulin being the most common as they indicate common signs of metabolic dysregulation (see Online Appendix 2).^{5,6,7,8,9,10,11,12,13} After the initial consultation, a clinic-standard 2-month follow-up is recommended with repeat labs drawn for any abnormal values. Because of this recommendation, patients may not have each laboratory value tested at every visit.

Descriptive statistics are reported as median with minimum and maximum or as mean and standard deviation for continuous variables depending on the data distribution. Categorical variables are presented as frequency and per cent. To account for variable patient follow-up time, frequency and the correlation of such repeated observations by individual patients, we estimated metabolic outcomes using separate linear mixed models with random intercepts. Additionally, patient age, and sex, as well as their separate two-way interactions with time, were estimated to evaluate the difference in change over the length of follow-up (up to 500 days after the first appointment). Finally, these models were re-estimated using a separate patient subset consisting of those with high initial HbA1C (values of 5.7% or greater). Alpha of 0.05 was used to determine statistical significance throughout.

This study was deemed exempt as secondary research for which consent is not required. The investigators conducted the review in a manner in which the identity of the participants cannot readily be ascertained. The principal investigator did not directly contact the participants, and all identifying information was coded and recycled after data analysis (Federal regulatory statute 45CFR46.104[d]). All procedures were approved by the local institutional review board (InfoEd record# 2005485).

Evaluating the change in HbA1c levels over time, overall unadjusted results estimate a small statistically significant decrease in HbA1c (−0.126% per 180 days, p < 0.0001). Additionally, age was positively associated with HbA1c (0.120% per 10-year increase, p < 0.0001).

Among patients with a high HbA1c baseline value, a statistically significant interaction was observed between age and follow-up days, suggesting that the rate of decrease was lower for older patients (interaction p = 0.0274, Figure 2).

Figure 2 shows the interaction between age (presented at 40 years, 50 years, and 60 years) and follow-up days within patients with a high HbA1c baseline value. For ease of interpretation, sex is excluded from the figure.

Evaluating the change in TG levels over time, overall adjusted results estimate a small statistically significant decrease (−8.262 mg/dL per 180 days, p = 0.0235). No other predictors were statistically significant. Within patients with a high HbA1c baseline value, no predictors were statistically significant, and no significant interactions were observed.

Evaluating the change in HDL-c levels over time, overall unadjusted results demonstrate a statistically significant increase in HDL-c (2.466 mg/dL per 180 days, p < 0.0001). Additionally, overall adjusted results demonstrated a statistically significant interaction between age and follow-up days, suggesting that the rate of increase associated with follow-up time may be driven by patient age (interaction p = 0.0347, Figure 3). Within patients with a high HbA1c baseline value, age (4.21 mg/dL per 10-year increase, p = 0.0215) and follow-up (3.618 mg/dL per 180 days, p < 0.001) but not sex (female 8.87 mg/dL higher, p = 0.0555) were significantly associated with HDL-c.

Figure 3 shows the interaction between age (presented at 40 years, 50 years, and 60 years) and follow-up days for the total patient sample. For ease of interpretation, sex is excluded from the figure.

Evaluating the change in hsCRP levels over time, overall adjusted results demonstrate a statistically significant decrease in hsCRP (−2.106 mg/L per 180 days, p < 0.0001). Additionally, age was a statistically significant negative predictor (−0.75 mg/L per 10-year increase, p = 0.0008). Overall, neither age nor sex demonstrated a significant interaction with follow-up time. Results among high HbA1c baseline patients mirrored the overall findings.

Evaluating the change in Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) levels over time, overall adjusted results demonstrate a statistically significant decrease in HOMA-IR (−0.553 per 180 days, p < 0.0001). Additionally, age was a statistically significant negative predictor (−0.55 per 10-year increase, p = 0.0001). Within patients with a high HbA1c baseline value, there was a statistically significant interaction between age and follow-up, suggesting a divergence in HOMA-IR by age, with older patients demonstrating a greater decrease and younger patients demonstrating a potential increase (interaction p = 0.0209, Figure 4).

Figure 4 shows the interaction between age (presented at 40 years, 50 years, and 60 years) and follow-up days within patients with a high HbA1c baseline value. For ease of interpretation, sex is excluded from the figure.

Both the full cohort and the high HbA1c group had a small expected decrease in TGs. The changes in the high HbA1c group were not great enough for statistical significance, however. The results require clinical interpretation to determine whether these small changes have any clinical benefit for the patient population. Given the evidence of risk associated with high TGs, small decreases may be clinically relevant.

The full cohort increased their HDL-c significantly in the study. The overall cohort showed a 10.2% expected 1-year increase in HDL-c from the initial mean. Comparatively, nicotinic acid, an approved medication for dyslipidaemia, may improve HDL-c by 10% – 30%. Nicotinic acid, however, reports side effects including cutaneous flushing, hepatotoxicity, insulin resistance and hyperuricaemia.¹⁴ This clinic is obtaining increases in the same range as medical intervention while avoiding the side effect profile. The high HbA1c group showed a larger 1-year increase of 14% from the initial, indicating further benefit in patients who are experiencing hyperglycaemia.

The expected reductions in hsCRP of the full cohort were 23.6% from initial, while the high HbA1c group had an expected decrease of 51.5%. Reductions in hsCRP on a long-term scale show that a focus on a VLCD may reduce systemic inflammation and thus reduce risk for future health complications. Given that hsCRP is a predictor of disease and may play a role in the disease process itself, any reductions can be viewed as beneficial.¹⁵

The high HbA1c group had a significant expected reduction in their HbA1c. Metformin glycinate, when used as monotherapy, reduces HbA1c in diabetic patients by around 1%.¹⁶ This clinic showed expected 1-year reductions like that of metformin (0.7%), the first-line treatment for glycaemic control in patients with type 2 diabetes. The patients in this clinic were not drug naïve, meaning they likely experienced pharmacologic benefits for their HbA1c before establishing care at the clinic. This indicates that care in a metabolic clinic, with a focus on a VLCD, may be a beneficial long-term addition to pharmacologic intervention for glycaemic control in pre-diabetic and diabetic patients.

Insulin resistance is the hallmark of type 2 diabetes, and the high HbA1c cohort had a 15% expected 1-year reduction in their HOMA-IR scores. By improving insulin sensitivity overall, this clinic is providing a long-term improvement for patients with diabetes. The expected changes in HOMA-IR were confounded by age. Older patients showed a larger benefit to their insulin sensitivity; however, younger patients showed a potential decline. Despite this, the overall decrease in HOMA-IR, signifying decreased insulin resistance, remained.

Patients with hyperlipidaemia, elevated HbA1c or insulin resistance likely presented to the clinic having already been prescribed the first-line medication for their disease state. Therefore, many patients had seen changes in HbA1c, lipids and other values directly because of their medications, leaving less room for improvement from other modalities. We argue that if this clinic (specifically, implementation of a VLCD) were their first-line treatment, before pharmacologic intervention, the expected changes may be even larger. Although not directly analysed in this study, the physicians do appreciate the ability to de-prescribe medications for their patients as metabolic parameters improve. The success of therapeutic carbohydrate reduction could be further supported by the de-prescription as a patient’s lifestyle improves. Further research is required to measure the impact of therapeutic carbohydrate reduction in medication-naïve patients, as well as properly analyse the rates of de-prescription in this clinic.

Clinical audits of this nature provide patient data from real healthcare strategies. Not only do these studies create the opportunity for quality improvement, but they can also reassure larger healthcare institutions that individual clinics are seeing positive outcomes in patients. This audit provides data that align with previous research on VLCDs, showing improved cardiovascular risk factors, insulin sensitivity and improved glucose markers.^3,4 By continuing to show the success of this implementation in clinical practice, we can encourage increased recommendations of a VLCD for the management of metabolic health.

This study has potential limitations. Firstly, conclusions about the positive effects in this study cannot be causally attributed to a VLCD itself because of the lack of a control group, random assignment or ability to measure compliance of patients to the diet. However, it is important to note that the focus of this metabolic health clinic is a VLCD, which differs significantly from many primary care models. Future research is needed to compare the results of this clinic to those of others that do not implement a VLCD. This would allow for a direct comparison of varying approaches to care.

Secondly, the limitation is the limited number of covariates. This study examined age and biological sex; however, body composition and body mass index could also be crucial components of a patient’s initial lab values and their progression over time. Lean versus obese individuals may show varying levels of change over time. Another limitation arises from this, in that this study does not control for weight loss over time. Therapeutic carbohydrate reduction in this clinic may drive weight loss, and the laboratory benefits in this clinic could be attributed to weight loss. Future research is needed to examine weight loss in this clinic and the associated biomarker changes.

As previously mentioned, this clinic does encourage lifestyle changes outside of solely nutrition, such as increases in physical activity and sleep. We acknowledge that these are crucial components of overall health, and real improvements may be attributed to patients changing these lifestyle habits upon presentation to the clinic. The focus of this research is on the nutritional aspects of lifestyle change, as this is the main focus of the providers and is discussed at the greatest length at each patient visit. We argue that therapeutic carbohydrate reduction is the largest driver of metabolic change in this patient population.

This study did not exclude patients with a medical history of mental health or eating disorders. This study was not designed with the intention of evaluating therapeutic carbohydrate reductions’ role specific to patients with these past diagnoses. Therefore, no conclusions can be drawn regarding the safety or effectiveness of this intervention in these groups.

Thirdly, we did not control for medication usage outside of exogenous insulin. Patients may have been on medication (such as metformin or atorvastatin) before coming to the clinic and may have started new medications while visiting the clinic. This creates another variable to which changes could be contributed to.